指纹识别技术作为应用最广泛的生物特征识别技术，已经普及于智能门锁、打卡机等电子设备的身份认证之上。随着指纹采集器变得更加小型化与轻便化，使得采集到的指纹图像越来越小，应用传统的指纹识别方法针对小面积指纹图像进行识别，会导致正确识别率大幅降低，同时LED灯作为光学指纹采集器使用最频繁的背光源，存在功耗大、背光不均匀等不足。因此，本文设计了一种应用PMOLED（Passive Matrix OLED）屏的光学指纹采集系统，并且从图像处理和深度学习两方面对采集的小面积指纹图像进行了识别方法研究。本文的主要工作如下：

（1）应用自发光PMOLED屏设计了一种光学指纹采集系统。针对该系统采集的原始图像存在屏幕发光像素干扰等问题，采用平均值消除法对原始数据进行了降噪处理，大幅地减少了屏幕干扰信号，之后进行了规格化、低通滤波、图像增强、二值化以及细化处理，得到了清晰的单像素小面积指纹图像。

（2）提出一种基于拓扑结构拼接的小面积指纹识别方法。该算法首先对细化后的指纹图像提取分叉点和端点两种细节特征点，求出每一个细节特征点的方向场角度，然后将两张指纹图像进行特征点匹配，对匹配的特征点进行拓扑结构连线，最后应用配准的特征点对之间的欧氏距离之和求取最优的旋转平移参数，实现小面积指纹拼接，提高正确接受率。

（3）提出一种基于改进残差网络融合卷积注意力机制的小面积指纹识别方法。将残差网络中的单一残差块改进为三种残差块类型，并通过不同的堆叠方式对残差块进行重新布局，减少了非线性激活函数过多带来的影响，然后将卷积注意力机制添加至改进的残差网络中，利用卷积注意力机制对所提取的指纹图像特征进行加权处理。该方法减少了网络训练和测试的收敛时间，同时提高了测试准确率。

实验结果表明，本文应用PMOLED设计的光学指纹采集系统可靠有效，并且本文提出的基于拓扑结构拼接的小面积指纹图像识别方法在自建库和重构后的FVC2002DB2指纹库分别取得99.57%和99.65%的正确识别率，基于改进残差网络融合卷积注意力机制的小面积指纹识别方法在自建库和重构后的FVC2002DB2指纹库分别取得98.15%和98.69%的测试准确率，证明了本文方法的有效性，满足智能产品的标准要求。

As the most widely used biometric identification technology, fingerprint identification technology has been widely used in the identification of intelligent door locks, punch cards and other electronic devices. As the fingerprint collector becomes more miniaturized and portable, the collected fingerprint image becomes smaller and smaller. The application of traditional fingerprint identification methods for small area fingerprint image recognition will lead to a significant reduction in correct recognition rate. At the same time, LED lights are the most frequently used backlight source for optical fingerprint collectors, which has the disadvantages of high power consumption and uneven backlight. Therefore, this paper designs an optical fingerprint acquisition system using PMOLED (passive matrix OLED) screen, and studies the recognition method of small area fingerprint image from image processing and deep learning. The main work of this paper is as follows:

(1) An optical fingerprint acquisition system based on self luminous PMOLED screen is designed. Aiming at the problems of the original image collected by the system, such as the interference of the screen luminous pixels, the average value elimination method is used to reduce the noise of the original data, which greatly reduces the screen interference signal. After normalization, low-pass filtering, image enhancement, binarization and refinement, the clear single pixel small area fingerprint image is obtained.

(2) A small area fingerprint recognition method based on topology splicing is proposed. Firstly, two kinds of minutiae feature points, bifurcation point and endpoint point, are extracted from the thinned fingerprint image, and the direction field angle of each minutiae feature point is calculated. Then, the two fingerprint images are matched by feature points, and the matching feature points are connected by topological structure. Finally, the optimal rotation and translation parameters are obtained by using the sum of Euclidean distances between matched feature points Product fingerprint stitching, improve the correct acceptance rate.

(3) A small area fingerprint identification method based on improved residual network and convolutional attention mechanism is proposed. The single residual block in the residual network is improved to three residual block types, and the residual blocks are rearranged by different stacking methods to reduce the impact of too many nonlinear activation functions. Then convolution attention mechanism is added to the improved residual network, and the extracted fingerprint image features are weighted by convolution attention mechanism. This method reduces the convergence time of network training and testing, and improves the accuracy of testing.

The experimental results show that the optical fingerprint acquisition system designed by PMOLED is reliable and effective, and the small area fingerprint image recognition method based on topology splicing achieves 99.57% and 99.65% correct recognition rates in the self built database and the reconstructed FVC2002DB2 fingerprint database, respectively.The small area fingerprint identification method based on improved residual network and convolutional attention mechanism achieves 98.15% and 98.69% test accuracy respectively in the self built database and the reconstructed FVC2002DB2 fingerprint database, which proves the effectiveness of this method and meets the standard requirements of intelligent products.

[1]游延筠, 邱文彬, 张龙, 等. 人工智能在智能门锁中的应用[J]. 人工智能, 2020(05): 58-67.

[2]白子轶, 毛懿荣, 王瑞平. 视频人脸识别进展综述[J]. 计算机科学, 2021, 48(03): 50-59.

[3]李明进. 指纹识别技术现状及发展趋势研究[J]. 无线互联科技, 2020, 17(01): 158-159.

[4]叶文炜, 何丽鹏, 林 峰. 指纹静脉多模识别光学系统的设计[J]. 激光与光电子学进展, 2018, 55(09): 370-376.

[5]周路, 迟耀丹, 郭亮. 物方远心双指指纹采集光学系统设计[J]. 激光与光电子学进展, 2016, 53(10): 214-219. 

[6]Arai T. Oxide‐TFT technologies for next‐generation AMOLED displays[J]. Journal of the Society for Information Display, 2012, 20(3): 156-161.

[7]Yan X, Tian C, Wei J, et al. A scalable seamless full color PMOLED[J]. Journal of the society for information display, 2014, 22(5): 245-250.

[8]刘中秋, 李健, 刘兆学, 等. 一种基于亮背景的光学指纹模块[P]. 中国专利: CN207380719U, 2018-05-18.

[9]叶金山, 刘国红. 一种指纹识别方法及移动终端[P]. 中国专利: CN107194331B, 2019-04-12.

[10]白王静, 李福山, 罗福根, 等. 侧入式LED背光源仿真中亮度均匀性的判定方法[J]. 电视技术, 2015, 39(7): 64-68.

[11]马东阁, 杨德志, 马远博. 等. 一种以有机发光二极管为背光源的光学指纹采集系统[P]. 中国专利: CN209842643U, 2019-12-24.

[12]吉紫娟, 包佳祺, 尹娟娟. 指纹及手指静脉图像复合采集光学系统设计[J]. 应用光学, 2017, 38(5): 770-776.

[13]吕岑, 朱欢敏. 基于结构光投影的三维指纹特征提取方法[J]. 科学技术与工程, 2014, 14(27): 71-74.

[14]Wang S, Zhang H, Zhang B, et al. Towards high-power-efficiency solution-processed OLEDs: Material and device perspectives[J]. Materials Science and Engineering R Reports, 2020, 140: 100547.

[15]邹 波. 新型LED背光源技术与应用实践研究[J]. 电子测试. 2017(16): 124-125.

[16]蒋洪在, 申亨哲, 尹一炫, 等. 感测PMOLED显示器中的照度的方法及其控制方法[P]. 中国专利: CN109579986A, 2019-04-05.

[17]罗志猛, 何秀凯, 赵云, 等. 一种PMOLED显示模组[P]. 中国专利: CN206471333U, 2017-09-05.

[18]Kyuragi H, Okazaki Y, Onishi S, et al. A Fingerprint Verification Algorithm Using the Differential Matching Rate[C]// Proceedings of the 16 th International Conference on Pattern Recognition (ICPR'02), 2002, 3: 799-802.

[19]Monika, Kumar M. A novel fingerprint minutiae matching using LBP[C]// International Conference on Reliability. IEEE, 2015: 1-4.

[20]Shabrina N, Isshiki T, Kuniena H. Fingerprint authentication on touch sensor using phase-only correlation method[C]// Proceedings of the 2016 7th International Conference of Information and Communication Technology for Embedded Systems. Piscataway NJ: IEEE, 2016: 85-89.

[21]Trunfio G A, Topa P, WS J. A new algorithm for adapting the configuration ofsubcomponents in large-scale optimization with cooperative coevolution[J]. InformationSciences, 2016, 372: 773-795.

[22]Zanganeh O, S Rinivasan B, Bhattacha Rjee N. Partial fingerprint matching through region-based similarity[C]// Proceed-ings of the 2014 International Conference on Digital Image Compu-ting: Techniques and Applications. Piscataway, NJ: IEEE, 2014: 1-8.

[23]Yamazaki M, LI D, ISSHIKI T, et al. SIFT-based algorithm for fingerprint authentication on smartphone[C]// Proceedings of the 2015 6th International Conference of Information and Communication Technology for Embedded Systems. Piscataway, NJ: IEEE, 2015: 1-5.

[24]Han X, Leung T, Jia Y, et al. MatchNet: unifying feature and metric learning for patch-based matching [C]// Proceedings of the 2015 Conference on Computer Vision and Pattern Recognition. Pis-cataway, NJ: IEEE, 2015: 3279-3286.

[25]郝培蓉. 基于分类特征的指纹识别系统的研究与实现[D]. 北京: 北京林业大学, 2020.

[26]汤正刚. 自动指纹识别系统中的关键技术研究[D]. 杭州: 杭州电子科技大学, 2019.

[27]王棪腾. 小面积指纹图像识别与匹配算法研究[D]. 西安: 西安电子科技大学, 2020.

[28]付青柳. 小面积指纹拼接算法研究[D]. 西安: 西安电子科技大学, 2017.

[29]范文博. 基于Android系统的小面积指纹识别技术的研究[D]. 西安: 西安科技大学, 2019.

[30]吴优. 基于双因子配准与脊线邻域矫正的小面积指纹拼接算法研究[D]. 西安: 西安电子科技大学, 2018.

[31]王宇萌. 面向移动终端的部分指纹拼接技术研究与实现[D]. 济南: 山东大学, 2018.

[32]Zhang F D, Feng J F. High-resolution mobile fingerprint matching via deep joint KNN-triplet embedding [C]// Proceedings of the 31st AAAI Conference on Artificial Intelligence. Menlo Park, CA: AAAI Press, 2017: 5019-5020.

[33]Kai C, Jain A K. Latent orientation field estimation via convolutional neural network[C]// 2015 International Conference on Biometrics (ICB). IEEE, 2015: 349-356.

[34]张永良, 周冰, 詹小四, 等. 基于深度学习的小面积指纹匹配方法[J]. 计算机应用, 2017(11): 3212-3218.

[35]钟瑞. 基于深度学习和多特征融合的小面积指纹识别算法研究[D]. 西安: 西安电子科技大学, 2018.

[36]陈文燕, 范文博, 杨钧宇. 基于密集连接卷积网络的小面积指纹识别方法[J]. 计算机工程与应用, 2020, 056(010): 134-140.

[37]李志鹏, 何晓彤, 胡金鑫, 等. 光学指纹采集与处理[J]. 物理实验, 2017, 37(z1): 4-8.

[38]王烈, 陈俊鸿, 罗文, 等. 指纹图像块方向准确性估计方法[P]. 中国专利: CN105740902B, 2019-06-11.

[39]王延年, 牛飞婷, 刘 婷. 基于Gabor滤波的指纹图像增强方法[J]. 激光杂志, 2016, 37(01): 117-119.

[40]Spiegel B M, Naliboff B, Mayer E, et al. Development and validation of a concise point-of-care severity index in IBS: The" BEST" questionnaire[C]//Gastroenterology. INDEPENDENCE SQUARE WEST CURTIS CENTER, STE 300, PHILADELPHIA, PA 19106-3399 USA: WB SAUNDERS CO-ELSEVIER INC, 2006, 130(4): A513-A513.

[41]Tico M, Kuosmanen P. Fingerprint matching using an orientation-based minutia descriptor[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2003, 25(8): 1009-1014.

[42]Feng J. Combining minutiae descriptors for fingerprint matching[J]. Pattern Recognition, 2008, 41(1): 342-352.

[43]刘晓杰, 张慧敏. 基于拓扑结构的图像拼接算法[J]. 电讯技术, 2017, 57(005): 555-560.

[44]肖刚, 高飞, 郭征, 等. 基于相位相关的指纹拼接方法研究[J]. 传感技术学报, 2005(04): 914-918.

[45]Kim Y D, Park E, Yoo S, et al. Compression of Deep Convolutional Neural Networks for Fast and Low Power Mobile Applications[J]. Computer Science, 2015, 71(2): 576-584.

[46]Wang Y , Xu C , You S , et al. CNNpack: packing convolutional neural networks in the frequency domain[C]// 29th Conference on Neural Information Processing Systems (NIPS 2016). IEEE, 2016, 41(10): 2495-2510.

[47]He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]// Proceed-ings of the2016 IEEE Conferen-ce on Computer Vision and Pattern Recognition. IEEE, 2016: 770-778.

[48]田强, 贾小宁. 基于深度残差网络的车标识别[J]. 吉林大学学报(理学版), 2021, 9(02): 319-324.

[49]马力, 帅仁俊, 刘文佳, 等. 基于改进的残差神经网络的白细胞分类[J]. 计算机工程与设计, 2020, 41(10): 2982-2987.

[50]唐风高, 伍雪冬. 一种改进的残差网络的人脸识别方法[J]. 计算机与数字工程, 2020, 48(09): 2248-2253.

[51]郭玥秀, 杨伟, 刘琦,等. 残差网络研究综述[J].计算机应用研究, 2020, 37(05): 1292-1297.

[52]赵志成, 罗泽, 王鹏彦, 等. 基于深度残差网络图像分类算法研究综述[J]. 计算机系统应用, 2020, 29(01): 14-21.

[53]Huang G, Sun Y, Liu Z, et al. Deep networks with stochastic depth[C]// Proceedings of the 14th European Conference on Computer Vision. Amsterdam, The Netherlands: Springer , 2016. 646-661.

[54]Zagoruyko S, Komodakis N. Wide Residual Networks[C]// British Machine Vision Conference, BMVC, 2016, 87: 1-12.

[55]He K, Zhang X, Ren S, et al. Identity Mappings in Deep Residual Networks[C]// Springer, Cham, 2016, 9908: 630-645.

[56]Woo S, Park J, Lee J Y, et al. CBAM: Convolutional Block Attention Module[C]// Computer Vision-ECCV, 2018, 11211: 3-19.

[57]陈虹旭, 李晓坤, 郑永亮, 等. 基于深度学习的指纹识别方法研究[J]. 智能计算机与应用, 2018, 8(03): 64-6.